Research On Personalized Shared Control Considering Driver's Driving Capability And Style

The shared control system can be defined as the safe,efficient,friendly and stable driving mode formed by overcoming the decision-making conflict between the driver with social attributes and the auto-driving system with logical attributes,which is the optimal driving match between the driver and the auto-driving system.Taking the social and ethical factors as well as the intelligent logic attributes of the auto-driving system into account,the shared control system could exist in the automated vehicle for a very long time.In shared control system,analysis of the coordination and conflicting mechanism between the human driver and the auto-driving system is the research foundation,and human factors as well as the decision-making logic of the auto-driving system are main factors affecting the system performance.Human factors are the pattern properties of driver's behaviors,such as the driving style,driving skill and driving status,and they're the key factors in the allocation of driving authority in the share control system.Besides,human factors have an important influence on the decision-making logic in auto-driving system.Therefore,research on the personalized shared control basing on the human factors is one of the key issues on the advancing intelligent transportation system.In this paper,basing on the analysis of the coordination and conflicting mechanism between the human driver and the auto-driving system in shared control system,the personalized shared control research considering driving capability and driving style is carried out,which focuses on the human factors and logical attribute of the auto-driving system.Theories and test methods of the system stimulate and scenario construction are established for the research on the human factor attributes.The definition and evaluation method of driving capability is proposed,aiming to improve the rationality for the driving authority allocation mechanism caused by the complex and changeable human factor attributes.The self-like auto-driving strategy is implemented basing on the characterization and evaluation method of self-like attribute in the intermediate degree.A high driver acceptability and safety shared control strategy is established basing on the shared controlframework consisting of the arbitration subsystem of driving authority and self-like auto-driving subsystem.Firstly,in order to meet the requirements for the characterization and evaluation of the human factors such as the driving capability and style for highly intelligent vehicles,theories and testing methods for the system stimulate and scenario construction is developed,as well as the driver-in-the-loop intelligent simulation platform and field data acquisition and validation platform.A V-shape testing process for human factor attributes is proposed.The stimulate signals and scenarios are selected basing on the periodicity and mutation of the scenario stimuli.The Random Vehicle-Road Field(RVRF)model and its corresponding micro driving scenarios those can reveal the human-vehicle-road coupling mechanism and vehicle-road cooperative rules are established by coupling the space topology of drivable area with the vehicle motion patterns.By establishing a complete system configuration and reasonable test procedures,the natural driving test method achieves a big data acquisition and test system with high precision,multidimensional and high scene consistency.Secondly,the concept and evaluation method of driving capability are proposed in order to improve the distribution rationality of driving authority in the shared control caused by the time-varying,high-order nonlinear and dynamic characteristics of human factor attributes.The driving capability is defined as the driver's gradient ability to control the vehicle with the change of the scenario load,which is a complex of human factors including driving style,driving skill,driving status,etc.and has time-varying nonlinear dynamic characteristics.The driving capability identification model is established basing on the Hammerstein model,and its key parameters are decoupled and dimensionalized by principal component analysis.The classification of driving capability is conducted by the combination of objective ant colony clustering and subjective scale analysis.The driving capability evaluation equation is obtained by multiple linear regression analysis.The attribute mechanism of driving capability is analyzed and its evaluation method is validated in typical stimulus scenarios and virtual RVRF.Thirdly,the characterization and evaluation framework for driving style corresponding to the intermediate self-like degree is established according to the characterization andevaluation requirements for the self-like attribute.The driving style is defined as a relatively stable and habitual internal behavioral tendency of drivers,and it's the combination of individual psychological thinking and behavior patterns which are with strong differences among different individuals.The characterization and evaluation system for the self-like attribute consists of the feature extraction,offline evaluation,online data arbitration and online evaluation methods of driving style.The classification database of the self-like attribute and its off-line identification model with optimization parameters are established by the objective-subjective combination based classification method,the multi-dimensional Gaussian hidden Markov process based identification model and the orthogonal optimization test based parameter optimization method.The online data arbitration is realized by the online traffic situation identification model considering the vehicle motion intention,and the online evaluation for the self-like attribute is realized by the online identification model of driving style.The evaluation accuracy and online identification cycle of the framework are verified and analyzed in the typical stimulate scenarios,RVRF scenarios and natural driving scenarios respectively.Finally,the personalized shared control strategy considering driving capability and style is developed.Coupling and cooperation modes between shared control system and other systems in the “driver-controller-scenario” system are analyzed,and the framework of the shared control system consisting of the driving authority arbitration subsystem and the self-like auto-driving subsystem is proposed.The driver and auto-driving system share the vehicle control by adjusting the driving authority allocation coefficient in real time during the dynamic driving task.A real-time driving authority arbitration mechanism is developed through the real-time driving capability identification model basing on multi-dimensional hybrid Gaussian recognition process.The self-like decision logic in the auto-driving system appropriate to the complex scenarios are realized basing on the two-layer hybrid Gaussian hidden Markov identification process and the hybrid observable Markov decision process.Complex scenarios including lanes,road topology and vehicle behaviors are built on the simulator platform and the field test platform,and the performance superiorities of the shared control system over human driver and auto-driving are verified and analyzed by theproposed evaluation criterion.